MI355X FP4: M3 EAGLE vs Off Speculative Decoding
Speculative decoding comparison of M3 EAGLE versus Off on MI355X FP4 (AMD CDNA 4) running MiniMax M3 428B. Throughput, cost, and interactivity differences across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.
MTP acceptance-rate implementations differ across inference engines. Points from different engines are not directly comparable on the same curve — throughput and cost at matched interactivity may reflect engine-level differences rather than pure speculative decoding gains. Interpret cross-engine comparisons with caution.
M3 EAGLE posts 3333 tok/s/chip for $0.13 per million tokens at 63 tok/s/user on MiniMax M3 428B (MI355X FP4); Off posts 1817 tok/s/chip for $0.23. M3 EAGLE is 83% cheaper per token; M3 EAGLE delivers 83% more tok/s/chip. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level.
Throughput at 99 tok/s/user on MiniMax M3 428B (MI355X FP4): M3 EAGLE hits 2365 tok/s/chip, Off hits 838. Per-million costs land at $0.18 and $0.50 respectively. M3 EAGLE is 182% cheaper per token; M3 EAGLE delivers 182% more tok/s/chip. Speculative decoding trades extra compute on draft tokens for fewer decoding steps — the payoff depends on sequence length and batch size.
Toward the upper edge of the 27–171 tok/s/user interactivity band, at 135 tok/s/user on MiniMax M3 428B (MI355X FP4): M3 EAGLE runs 1702 tok/s/chip at $0.24/M tokens, Off runs 324 at $1.29/M. M3 EAGLE is 425% cheaper per token; M3 EAGLE delivers 425% more tok/s/chip. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less. (Numbers reflect this URL's pinned 1k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | M3 EAGLE:3332.8Off:1817.1 | M3 EAGLE:2364.9Off:837.7 | M3 EAGLE:1701.7Off:324.2 |
| Cost ($/M tok) | M3 EAGLE:$0.125Off:$0.229 | M3 EAGLE:$0.176Off:$0.497 | M3 EAGLE:$0.245Off:$1.285 |
| tok/s/MW | M3 EAGLE:1594650Off:869404 | M3 EAGLE:1131546Off:400801 | M3 EAGLE:814219Off:155098 |
| Concurrency | M3 EAGLE:~111Off:~58 | M3 EAGLE:~51Off:~17 | M3 EAGLE:~27Off:~5 |
Inference Performance
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.